Review Request: Statistics Synchronization: final review request

[Edward Hades Toroshchin]

On Mon, Nov 19, 2012 at 09:11:18PM +0100, Matěj Laitl wrote:
> > 
> > You have this restriction in StatSync::Provider, and according to
> > yourself.

I seem to have swallowed the end of the sentence. It should have gone
like this:

...according to yourself, existing QueryMakers won't work unless created
in the main thread.

> This is why I'm reluctant to author BlockingQueryMaker, as it would have 
> either a sloppy API (with restrictive constraints) or would be tricky to 
> implement correctly *in my humble opinion*.

I see your point. You mean that any user of BlockingQueryMaker would
also inherit its curse^Wthread restriction. All right, let's have it
this way for now.

> On the other hand, if someone else (you?) wants to create it (perhaps reusing 
> some StatSyncing::Provider code) I'm happy to review the code and use it 
> instead of ad-hoc code in Provider once it proves working correctly.

I'd rather rewrite the whole deal, being extremely careful around
threads, so that in the end we don't have to run stuff in the main
thread, just because we don't know what could happen otherwise.

> P.S.: Edward, when you've already compiled the StatSyncing code, perhaps you 
> can lightly test it? ;) UI in Amarok Config, scrobbling still working, 
> synchronizing with Last.fm (warning: non-interactive part is time-consuming) 
> or with an iPod...

Sure, just don't wait for me before merging, if I find anything, we can
fix it already in the main tree.

-- 
Edward "Hades" Toroshchin
dr_lepper on irc.freenode.org